REQUIRED GUARD BANDS FOR CO-OPERATION OF DVB-T AND UMTS

Transcription

1 REQUIRED GUARD BANDS FOR CO-OPERATION OF DVB-T AND UMTS Christian Hamacher ComNets-RWTH Aachen, Kopernikusstr. 16, D-5274 Aachen, Germany, Abstract - In the IST DRiVE project, a flexible, hybrid communication system has been developed which performs dynamic allocation of spectral resources from a common region of spectrum to the participating radio systems. Thus, new spectral neighborhoods emerge, which pose requirements regarding the protection of one system from the adjacent channel interference generated by the other system. As an example, DRiVE has focused on a hybrid system consisting of digital TV broadcast (DVB-T) and UMTS. The resulting spectral neighborhood of powerful DVB-T transmitters and UMTS requires guard bands between the dynamically assigned spectral regions, and thus decreases the overall spectral efficiency of the hybrid radio system. This paper presents the results of simulative studies on the required size of the guard band in order to provide sufficient QoS in the participating systems. Additionally, the importance on the implementation-dependent transmitter and receiver quality is examined, taking into account laboratory measurement of available DVB transmitters and receivers. It is shown that while worst case requirements on DVB transmitter out of band performance set forth in the relevant standard can lead to poor system performance in some scenarios, state-of-the art equipment can easily exceed those requirements, and suitable allocation of channels can further improve the system performance, thus leading to satisfactory coexisting operation of UMTS and DVB-T, with only a small loss of spectral efficiency due to guard bands. Keywords UMTS, DVB-T, Coexistence, Hybrid Networks I. INTRODUCTION Given the well known constraints of high cost and limited availability of radio spectrum, efficient spectrum usage is key to the economic success of third generation cellular systems. The DRiVE IST project presented the novel concept of using available resources more efficiently by combining existing radio systems into a coordinated, hybrid system, in order to attain maximum flexibility in using available resources. The DRiVE system offers the user an integrated transport system which can adapt to both per-session QoS demands and overall system load status by dynamically choosing the most suitable RAN for the request, and by dynamically allocating spectral resources to the participating systems (Dynamic Spectrum Allocation, DSA) [1]. In the hybrid system outlined above, due to DSA, DVB-T and UMTS can be spectral neighbors, and therefore any outof-band energy emitted by one system will interfere with the adjacent system. Figure 1 shows such an Adjacent Channel Interference (ACI) scenario. The transmitter (TX) masks are defined in the relevant standards for DVB-T and UMTS [2][3][4]. However, the receiver filters (RX) are not standardized, and are here assumed to be ideal rect. Figure 1: Interference due to mask overlap ACI degrades the Signal to Noise Ratio (C/I) in the victim system, and therefore potentially decreases system capacity and Grade of Service (GoS) in the victim system. Therefore, guard bands (GB) are used between the spectral regions assigned to the interfering systems. In this context, GB is assumed to be the gap between the upper spectral border in the channel raster of one system, and the lower spectral border of the second system s channel raster. Since Guard Bands are unusable in either of the neighboring systems, they degrade overall spectral efficiency. Knowledge of the required guard bands is therefore necessary to estimate the benefit of a DRiVE system over traditional, fixed spectrum system concepts. II. SIMULATION PARAMETERS A. Parameters and Assumptions A detailed description of the simulation tool used for these studies can be found in [5]. Since the focus of DRiVE was on vehicular, high speed mobility, a typical highway scenario has been chosen. Path loss is assumed to be governed by an Okumura-Hata suburban model. Channel and mobility modeling are done with a temporal resolution of 1ms. Based on transmitter masks and receiver filters according to the relevant standards, interference levels are calculated /2/$ IEEE PIMRC 22

2 For computational efficiency, the size of the scenario has to be restricted to seven UMTS cells in a hexagonal grid (1 km radius), and one DVB-T cell (2 km radius). Only the central UMTS cell is evaluated, the outer ring serves as a source of intra-system interference and to realistically model handover. The UMTS cluster and an underlying road map can be positioned at different locations relative to the DVB-T transmitter, in order to capture the various interference situations from co-sited UMTS/DVB-T transmitters to border situation with handover between two DVB-T cells. Under the assumption of a downlink capacity shortage due to asymmetric traffic, we have focused on interference between a UMTS downlink and a neighboring DVB-T downlink. Both systems are assumed to have omnidirectional antennae. For the UMTS system, power control is C/I-based, with a CIR target for the 12.2kbit/s speech service of 17.75dB for both up- and downlink, and an abrupt RLF limit of 21dB. Vehicles are assumed to travel on a road grid. Since the choice of a roadmap is completely arbitrary, we chose a rectangular grid with turn probabilities according to the Manhattan grid described in [6]. The map covers the whole UMTS cell cluster, with a terminal velocity of km/h. A user is deemed to be satisfied if his call completes successfully. B. Simulator Parameterization In a baseline simulation series, the load of 45 Erlang in the uninterfered UMTS system has been determined at which a user satisfaction of 98% is attained. All subsequent simulations have been performed with the same load. For a series of initial worst case simulations, DVB transmission power for a cell radius of 2km has been estimated at dbm on the basis of a Hata-Okumura path loss model for the small town suburban case, a required C/I of 11dB for the 16-QAM modulation scheme, an assumed shadowing variance of 1dB and an additional safety margin. For the simulations presented in section IV, this power has been lowered to 74dBm, since the measurements described in section III showed a high resistance of the COFDM transmission scheme against adjacent channel interference, and thus a lower required minimum C/I at the DVB cell border. Figure 2 shows the results in terms of the percentage of satisfied users (USC) for the UMTS downlink adjacent to the interfering DVB link, and for a location at the center and the border of the interfering DVB cell. While at the border of the DVB cell, the interference power has decreased sufficiently to not degrade UMTS operation, in the center of the DVB cell provision of a guard band is not suitable to ensure satisfactory operation. Instead, strategies for increasing the average wanted signal level at the cell border (e.g. shrinking the cell around the DVB interferer) promise better success without excessive guard bands. Section IV.B gives a first estimate of the impact of modifying the size of the UMTS cell containing the DVB interferer. Figure 2: Worst case results The worst case results presented above are based on several pessimistic assumptions: first, the transmission mask assumed for the DVB transmitter has been chosen to be the envelope of the worst permissible masks according to [2]. Section III gives arguments for a more optimistic spectral mask. Additionally, Figure 3 shows that max. per link BS power and shadowing variance have a strong influence on the predicted GoS in the UMTS system. Both parameter have been chosen pessimistically for the initial worst case simulations guard band [MHz] Shadowing Variance [db] DVB BS distance [km] DVB cell center DVB cell border DL max TX Power [dbm] Figure 3: Dependence of UMTS user satisfaction on various simulation parameters Based on the suburban motorway scenario, reducing the shadowing variance to 7dB seems appropriate. Also, for the initial simulations, a maximum per-link power of 33dBm had been assumed for the UMTS BS. However, Figure 3 shows that the maximum available power per link has significant influence on the overall system performance. In a real system, power would be limited by a per-bs power budget rather than per individual link. Therefore, for the results in section IV, the simulation model has been refined to assume a per-bs power budget of 43dBm. Finally, the last

3 graph in Figure 3 indicates that the strong deterioration of UMTS performance is restricted to a rather small area around the DVB transmitter. DVB spectrum mask III. DRIVE MEASUREMENTS Within the DRiVE project, protection ratio measurements have been performed using market-available DVB transmitters and receivers. Based on these measurements, modified protection requirements have been obtained, leading both to modified masks and power levels for a second set of simulations. Measurement results presented in [5] indicate that the worst case assumptions made in section II.B are overly pessimistic with regard to both the required C/I for sufficient DVB performance and thus the required DVB transmission power, and also regarding the shape of the transmission mask. The DVB transmitters used for the DRiVE measurements represent those available for operation of the BBC DVB-T network operational in the UK. Those transmitters do exceed the Critical Cases mask from [2] in some places, but correspond reasonably well looking at integrated out of channel power, which is the only criterion relevant for the simulations. The Critical Cases mask represents a considerable improvement over the worst case mask used for the first results. However, the measurements at the same time illustrated that receiver nonlinearity, which is not covered in the simulation model at all, also plays an important role in overall system performance, somewhat degrading the improvement gained from using the better masks. Therefore, in [5] a modified transmission mask has been deduced, designed to include the nonlinear degradation caused by imperfect receivers into the transmitter signal. This was achieved by first identifying intermodulation products (IP) as the most important source of receiver degradation. Then, the overly optimistic Critical Cases mask was deteriorated by adding intermodulation shoulders at a variable level. The IP level was now adjusted until the resulting theoretical protection ratio curves agreed with those obtained by laboratory measurements of available receivers. Thus, a transmission mask was estimated that still allowed interference calculation by linear operations based on masks and filters, but included knowledge of the state of the art transmitter masks and the degrading non-linear effects in the receivers. It should be noted that within DRiVE, no UMTS receivers were available for measurements during the project duration. Therefore, receiver generated degradation was measured for DVB receivers, and the assumption was made that UMTS receivers will exhibit similar behavior. Additionally, these effects will depend on the power of the received signal, while in our simulations a constant deterioration of the transmitter mask was used. However, the measurements can still serve as an indication that a transmission mask between the best- and worst-case masks from the standard should be used. ACL [db] C/I (db) MHz Figure 4: DVB transmission masks guard band [MHz] Figure 5: ACL of DVB masks Figure 4 shows the worst case, best case and the modified DVB masks. From these masks, the Adjacent Channel Leakage (ACL) for a UMTS channel next to the DVB channel can be calculated as a function of the guard band between the two channels. The results of this calculation is presented in Figure 5, indicating a significant improvement over the worst case of the adjacent channel suppression even with the degraded DVB mask. IV. RESULTS Worst Case DRiVE mask Best case `Worst Case` mask DRiVE Mask `Critical Case` mask 3 4 A. Modified Spectrum Mask Figure 6 shows the improved predicted system performance for the changed parameters. For the left graph, the worst case mask was used, while for the right graph, in addition to the refined modelling, the modified DVB transmitter spectral mask was used. Outside a region of about one UMTS cell size, acceptable performance is attainable with a guard band of smaller than 1MHz width. Using the modified mask, a guard band of 5kHz leads to acceptable user satisfaction in the UMTS system outside a region of approx.7km around the DVB interferer. The loss in spectral efficiency caused by e.g. blocking the complete

4 adjacent channel in one UMTS cell per DVB cell is far smaller than the loss incurred by providing a wide guard band over the whole area of the DVB cell d=.7km d=1km,,5 1, Guard Band [MHz] 9 85 d=,5km 75 d=,7km 7,,5 1, Guard Band [MHz] Figure 6: Results using modified parameters (left), and additionally modified DVB mask (right) B. Cell Shrinkage As can be seen in Figure 5, the ACL of the different masks has a rather shallow slope with increasing guard band size. This fact manifests as a weak dependence of the attained UMTS performance on the size of the guard band, once the initial steep flank of the out-of-band emission has decayed. usable spectrum. Due to the large difference in cell size between the two coexisting radio systems, such a local tradeoff can greatly improve overall spectral efficiency, since only a fraction of all UMTS cells is affected. C. Impact on the DVB system So far, we have looked at the performance of the UMTS system, which can readily be evaluated looking at the USC. For the DVB downlink, such a criterion is less easy to obtain. While [6] defines a USC for packet based links, it is based on packet throughput, which is largely dependent on higher layer protocols, and thus difficult to model in a link level simulator. However, looking at Figure 8 and Figure 9, it becomes clear that no performance problems need to be expected on the DVB side: interference is close to background noise at the border, where DVB signal levels are low and therefore UMTS power control limits the BS transmission power. In the center of the DVB cell, where due to high DVB interference UMTS power control reaches the BS power limit, the extremely high DVB wanted signal level guarantees favorable C/I conditions for DVB, while UMTS performance is the system bottleneck. It therefore seems reasonable, instead of trying to further minimize the ACI by increasing the guard band, to increase the wanted signal level, e.g. by shrinking the UMTS cell containing the DVB interferer. At the expense of additional infrastructure once per DVB cell, one can thus avoid the decrease of spectral efficiency over the whole area associated with the general provision of a wider guard band. For the simulation results presented in Figure 7, the evaluated center cell containing the DVB transmitter at a distance of 5m to the victim UMTS BS was shrunk by moving the surrounding BS closer. All other parameters were the same as above. Figure 8: DVB CIR, border GB,5MHz / d=,5km,7,8,9 1, UMTS cell radius [km] Figure 7: Improved performance using smaller cell size The graph clearly shows the improved GoS due to the increased available receiver power at the cell border. As opposed to an increase of the guard band size, this strategy represents a local tradeoff of infrastrucure expense versus Figure 9: DVB CIR, center

5 V. CONCLUSIONS The studies performed in DRiVE have shown that coexistence of DVB-T and UMTS in a DRiVE hybrid network is possible. While the high power of the DVB transmitter creates considerable adjacent channel interference in a neighbouring UMTS channel, with a combination of guard band and minimum spatial separationof the DVB transmitter and victim UMTS cells occupying the adjacent channel a satisfactory grade of service can be obtained. In a series of simulations, estimates of minimum required guard bands have been determined, in order to reach satisfactory GoS in the UMTS system. The results of these simulations were shown to be especially sensitive to the shape of the DVB transmission mask. As-suming a worst case transmission mask as permitted by the relevant DVB standard, the simulations predict a severe capacity loss in the UMTS system, even with an extremely wide guard band. However, the results of the DRiVE laboratory measurements indicate that state of the art DVB transmitters perform much better than required by the standard they almost reach the performance of the demanding Critical Cases mask also defined in the DVB standard. Given such rather clean transmitters, the measurements also indicate that neglecting the influence of receiver imperfections is no longer possible. Therefore, a modified, degraded transmission mask for the DVB system has been chosen, which tries to account for receiver imperfections by artificially degrading the transmitter performance. Using this modified mask, simulation results predict that satisfactory performance of the critical UMTS system can be attained at a guard band of approximately 5kHz over almost the whole DVB cell area. Only an area of approximately one UMTS cell around the DVB interferer will experience stronger capacity loss, and therefore has only restricted access to the frequency channel adjacent to the DVB transmitter. However, since this degraded area is small compared to the total area of the DVB cell, the loss of spectral efficiency caused by DVB interference is very small. While the impact of DVB interference on UMTS can be significant, simulation results indicate that UMTS adjacent channel interference will have a negligible effect on DVB. Mainly due to the availability of power control in the UMTS system and the high robustness of DVB against ACI on the one hand side, and the high power of the DVB transmitters on the other side, interference into DVB is not the limiting factor for overall system performance in a DRiVE hybrid radio system. Results from the DRiVE coexistence studies have identified several fields for further research, in order to obtain more detailed and accurate predictions of spectral efficiency of the DRiVE system. First, since no UMTS receivers were available for measurement, receiver imperfections found in the DVB receivers have been assumed to be similarly present in the UMTS receivers. Once UMTS receivers are available, additional laboratory measurements covering DVB interference into UMTS would be beneficial to confirm this assumption. Secondly, once a more detailed analysis of the tradeoff between reduction of the cell size of the UMTS cells closest to the DVB interferer has been performed, based on this cell size and the estimated guard bands, an estimate of the overall spectral efficiency of the hybrid system can be attempted. Also, the measurement results have shown that the commonly used description of interference by linear integration of one system s adjacent channel leakage into the other system s passband is not accurate, since non-linear effects in the receivers can dominate the performance in the case of relatively clean transmitters. Therefore, the mask and filterbased approach of modelling adjacent channel interference, though commonly used in system evaluation, needs further refinement. AKNOWLEDGEMENTS This work has been performed in the framework of the IST project IST DRiVE, which is partly funded by the European Union, and will continue in the followup project IST overdrive. The DRiVE consortium consists of Ericsson (co-ordinator) BBC, Bertelsmann, Bosch, DaimlerChrysler, Nokia, Tecsi, Teracom, VCON, and Vodafone as well as Rheinisch-Westfälische Technische Hochschule RWTH Aachen, Universität Bonn, Heinrich- Hertz-Institut Berlin and the University of Surrey. The authors acknowledge the contributions of their colleagues in the DRiVE consortium. REFERENCES [1] J. Huschke, P. Leaves, Dynamic Spectrum Allocation Algorithm including Results of DSA Performance Simulations, IST DRiVE deliverable D9, Jan. 22 [2] ETS 3 744, Digital Video Broadcasting (DVB); Framing structure, channel coding and modulation for digital Terrestrial television (DVB-T) [3] 3GPP TS 25.11, UE Radio Transmission and Reception (FDD) [4] 3GPP TS 25.14, UTRA (BS) FDD; Radio transmission and Reception [5] Ch. Hamacher, J. Salter, Coexistence Measurement and Simulation Results, DRiVE deliverable D1, Mar. 22 [6] 3GPP Selection Procedures for the Choice of Radio Transmission Technologies for the UMTS (UMTS 3.3), April 1998